Analysis
Resid(COVID) + Joe Choo, Glen Morgenstern, Carrie Wang, and Zhixue (Mary) Wang
Research Question
The coronavirus has hit perhaps no American city quite as hard as New York, where more than 10,000 residents have died due to it. Yet, New Yorkers have heard conflicting advice from Governer Andrew Cuomo and President Donald Trump. Our team decided to look at how the stay-at-home order and other advice from government officials affected how New Yorkers drive. With reduced traffic, are accidents down? Has the lockdown affected New York’s six boroughs differently?
Motivating question: How has government guidance affected New York drivers’ motivations and safety?
Data
The data comes from the City of New York’s website. This data set consists of police reports on all motor vehicle collisions in New York City as of April 18. A police report is warranted when there is a fatality or injury or damage of $1000 or more. Our analysis will focus only on collisions in 2020.
You can find the updated data set at: https://data.cityofnewyork.us/Public-Safety/Motor-Vehicle-Collisions-Crashes/h9gi-nx95?fbclid=IwAR1UoErhqzmJrviRZ4zkpYu7cHhOgCAA417o6A3rhrIZSKrXPPNVjdzsWUQ
Set Up
library(tidyverse)## ── Attaching packages ─────────────────────────────────────────────────────────────────────────────────────────────────── tidyverse 1.2.1 ──## ✓ tibble 3.0.0 ✓ purrr 0.3.3
## ✓ tidyr 1.0.2.9000 ✓ dplyr 0.8.5
## ✓ readr 1.1.1 ✓ forcats 0.3.0## ── Conflicts ────────────────────────────────────────────────────────────────────────────────────────────────────── tidyverse_conflicts() ──
## x lubridate::as.difftime() masks base::as.difftime()
## x lubridate::date() masks base::date()
## x dplyr::filter() masks stats::filter()
## x readr::guess_encoding() masks rvest::guess_encoding()
## x lubridate::intersect() masks base::intersect()
## x dplyr::lag() masks stats::lag()
## x purrr::pluck() masks rvest::pluck()
## x lubridate::setdiff() masks base::setdiff()
## x lubridate::union() masks base::union()library(lubridate)data <- read.csv("data/Motor_Vehicle_Collisions_-_Crashes.csv")Dates to look at: Jan 20 - First reported US case Feb 9 - Death toll in China surpasses SARS epidemic (811 deaths) Feb 29 - First death on American soil March 1 - First confirmed case in NYC March 7 - Cuomo declares state of emergency March 11 - WHO declares pandemic, Trump bans travel from 26 European countries March 13 - Trump declares national emergency March 14 - First coronavirus death in NYC March 16 - NYC schools close (ordered by Cuomo) March 17 - De Blasio says NYC considering order, Cuomo says it won’t happen; schools, bar, restaurants close in NYC (de Blasio) March 20 - Cuomo orders lockdown March 22 - Lockdown starts (8pm) March 28 - Trump signs $2.2 trillion stimulus bill April 3 - CDC recommends wearing masks April 6 - Cuomo extends stay at home order to April 29 April 14 - Cuomo says he would defy Trump order to reopen New York April 16 - Cuomo extends stay at home order to May 15 April 17 - Trump tweets “Liberate” Minnesota, Michigan, Virginia
Data Wrangling
Make all street names upper case.
data$ON.STREET.NAME = toupper(data$ON.STREET.NAME)
data$OFF.STREET.NAME = toupper(data$OFF.STREET.NAME)
data$CROSS.STREET.NAME = toupper(data$CROSS.STREET.NAME)Confirm no duplicate collision IDs.
data[duplicated(data$COLLISION_ID), ]## [1] X CRASH.DATE
## [3] CRASH.TIME BOROUGH
## [5] ZIP.CODE LATITUDE
## [7] LONGITUDE LOCATION
## [9] ON.STREET.NAME CROSS.STREET.NAME
## [11] OFF.STREET.NAME NUMBER.OF.PERSONS.INJURED
## [13] NUMBER.OF.PERSONS.KILLED NUMBER.OF.PEDESTRIANS.INJURED
## [15] NUMBER.OF.PEDESTRIANS.KILLED NUMBER.OF.CYCLIST.INJURED
## [17] NUMBER.OF.CYCLIST.KILLED NUMBER.OF.MOTORIST.INJURED
## [19] NUMBER.OF.MOTORIST.KILLED CONTRIBUTING.FACTOR.VEHICLE.1
## [21] CONTRIBUTING.FACTOR.VEHICLE.2 CONTRIBUTING.FACTOR.VEHICLE.3
## [23] CONTRIBUTING.FACTOR.VEHICLE.4 CONTRIBUTING.FACTOR.VEHICLE.5
## [25] COLLISION_ID VEHICLE.TYPE.CODE.1
## [27] VEHICLE.TYPE.CODE.2 VEHICLE.TYPE.CODE.3
## [29] VEHICLE.TYPE.CODE.4 VEHICLE.TYPE.CODE.5
## <0 rows> (or 0-length row.names)No duplicates—all good.
data <- data %>%
select(-X)data %>%
count(CONTRIBUTING.FACTOR.VEHICLE.1)## # A tibble: 56 x 2
## CONTRIBUTING.FACTOR.VEHICLE.1 n
## <fct> <int>
## 1 "" 160
## 2 "Accelerator Defective" 22
## 3 "Aggressive Driving/Road Rage" 226
## 4 "Alcohol Involvement" 486
## 5 "Animals Action" 45
## 6 "Backing Unsafely" 1723
## 7 "Brakes Defective" 152
## 8 "Cell Phone (hand-Held)" 16
## 9 "Cell Phone (hands-free)" 1
## 10 "Driver Inattention/Distraction" 10710
## # … with 46 more rowsRegroup so that accerlator defective, brakes defective, windshield inadequate, tire failure, tow hitch defective, headlights defective, and other lighting defects are in one group called mechanical defect; using on board navigation device, texting, cell phone (hand-held), cell phone (hands-free), and other electronic device are under distraction; listening is under distracted; fatigued/drowsy, fell asleep, and lost consciousness are under drowsy; pavement defective, shoulders defective, and pavement slippery under environment.
data <- data %>%
mutate(CONTRIBUTING.FACTOR.VEHICLE.1 = as.character(CONTRIBUTING.FACTOR.VEHICLE.1)) %>%
mutate(CONTRIBUTING.FACTOR.VEHICLE.1 = case_when(
CONTRIBUTING.FACTOR.VEHICLE.1 == "" ~ "Unspecified",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Glare" ~ "View Obstructed/Limited",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Tire Failure/Inadequate" ~ "Mechanical Defect",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Windshield Inadequate" ~ "Mechanical Defect",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Tow Hitch Defective" ~ "Mechanical Defect",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Accelerator Defective" ~ "Mechanical Defect",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Brakes Defective" ~ "Mechanical Defect",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Headlights Defective" ~ "Mechanical Defect",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Other Lighting Defects" ~ "Mechanical Defect",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Cell Phone (hand-Held)" ~ "Driver Inattention/Distraction",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Listening/Using Headphones" ~ "Driver Inattention/Distraction",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Other Electronic Device" ~ "Driver Inattention/Distraction",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Using On Board Navigation Device" ~ "Driver Inattention/Distraction",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Cell Phone (hands-free)" ~ "Driver Inattention/Distraction",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Outside Car Distraction" ~ "Driver Inattention/Distraction",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Passenger Distraction" ~ "Driver Inattention/Distraction",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Texting" ~ "Driver Inattention/Distraction",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Eating or Drinking" ~ "Driver Inattention/Distraction",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Fatigued/Drowsy" ~ "Tired",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Fell Asleep" ~ "Tired",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Lost Consciousness" ~ "Tired",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Lane Marking Improper/Inadequate" ~ "Road Environment",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Obstruction/Debris" ~ "Road Environment",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Pavement Defective" ~ "Road Environment",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Pavement Slippery" ~ "Road Environment",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Shoulders Defective/Improper" ~ "Road Environment",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Prescription Medication" ~ "Health",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Physical Disability" ~ "Health",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Illnes" ~ "Health",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Traffic Control Device Improper/Non-Working" ~ "Road Environment",
CONTRIBUTING.FACTOR.VEHICLE.1 == "Tinted Windows" ~ "View Obstructed/Limited",
TRUE ~ CONTRIBUTING.FACTOR.VEHICLE.1
))data %>%
count(CONTRIBUTING.FACTOR.VEHICLE.1)## # A tibble: 29 x 2
## CONTRIBUTING.FACTOR.VEHICLE.1 n
## <chr> <int>
## 1 Aggressive Driving/Road Rage 226
## 2 Alcohol Involvement 486
## 3 Animals Action 45
## 4 Backing Unsafely 1723
## 5 Driver Inattention/Distraction 10915
## 6 Driver Inexperience 618
## 7 Driverless/Runaway Vehicle 33
## 8 Drugs (illegal) 21
## 9 Failure to Keep Right 48
## 10 Failure to Yield Right-of-Way 2876
## # … with 19 more rowsdata %>%
count(VEHICLE.TYPE.CODE.1)## # A tibble: 141 x 2
## VEHICLE.TYPE.CODE.1 n
## <fct> <int>
## 1 "" 324
## 2 "2 dr sedan" 8
## 3 "3-Door" 5
## 4 "4 dr sedan" 117
## 5 "AMB" 3
## 6 "AMBU" 1
## 7 "AMBUL" 38
## 8 "Amb" 1
## 9 "Ambul" 7
## 10 "Ambulance" 144
## # … with 131 more rowsRegroup so EMERG, AMB, AMBU, AMBUL, amb, ambul are under ambulance; BOX T is under Box Truck; DELIV, DELV, deliv, devli, Deliv are under delivery; DUMP, dump is under dump; FIRE, Fire, Firet, fire, and FIRET and under Firetruck; FLATB is under Flat Bed; FORK and FORKL is under Forlift; PICKU, pick, and Pickup with mounted is under Pick-up Truck; SCHOO, schoo, Schoo under School Bus; all wheeler sednas under Sedan; Tow, tow truck, tow t, under tow truck; UTIL, UTILI, Utili under utility
data <- data %>%
mutate(VEHICLE.TYPE.CODE.1 = as.character(VEHICLE.TYPE.CODE.1)) %>%
mutate(VEHICLE.TYPE.CODE.1 = case_when(
VEHICLE.TYPE.CODE.1 == "van" ~ "Van",
VEHICLE.TYPE.CODE.1 == "UTIL" ~ "Utility",
VEHICLE.TYPE.CODE.1 == "UTILI" ~ "Utility",
VEHICLE.TYPE.CODE.1 == "Utili" ~ "Utility",
VEHICLE.TYPE.CODE.1 == "" ~ "Unknown",
VEHICLE.TYPE.CODE.1 == "TOW T" ~ "Tow Truck",
VEHICLE.TYPE.CODE.1 == "Tow" ~ "Tow Truck",
VEHICLE.TYPE.CODE.1 == "Tow t" ~ "Tow Truck",
VEHICLE.TYPE.CODE.1 == "Tow Truck / Wrecker" ~ "Tow Truck",
VEHICLE.TYPE.CODE.1 == "TRK" ~ "Truck",
VEHICLE.TYPE.CODE.1 == "TRUCK" ~ "Truck",
VEHICLE.TYPE.CODE.1 == "2 dr sedan" ~ "Sedan",
VEHICLE.TYPE.CODE.1 == "3-Door" ~ "Sedan",
VEHICLE.TYPE.CODE.1 == "4 dr sedan" ~ "Sedan",
VEHICLE.TYPE.CODE.1 == "REFG" ~ "Refrigerated Van",
VEHICLE.TYPE.CODE.1 == "pick" ~ "Pick-up Truck",
VEHICLE.TYPE.CODE.1 == "PICKU" ~ "Pick-up Truck",
VEHICLE.TYPE.CODE.1 == "Pickup with mounted Camper" ~ "Pick-up Truck",
VEHICLE.TYPE.CODE.1 == "FORKL" ~ "Forklift",
VEHICLE.TYPE.CODE.1 == "FORK" ~ "Forklift",
VEHICLE.TYPE.CODE.1 == "FLATB" ~ "Flat Bed",
VEHICLE.TYPE.CODE.1 == "TRAC" ~ "Tractor Truck",
VEHICLE.TYPE.CODE.1 == "TRACT" ~ "Tractor Truck",
VEHICLE.TYPE.CODE.1 == "Tract" ~ "Tractor Truck",
VEHICLE.TYPE.CODE.1 == "tract" ~ "Tractor Truck",
VEHICLE.TYPE.CODE.1 == "Tractor Truck Diesel" ~ "Tractor Truck",
VEHICLE.TYPE.CODE.1 == "Tractor Truck Gasoline" ~ "Tractor Truck",
VEHICLE.TYPE.CODE.1 == "SCHOO" ~ "School Bus",
VEHICLE.TYPE.CODE.1 == "schoo" ~ "School Bus",
VEHICLE.TYPE.CODE.1 == "Schoo" ~ "School Bus",
VEHICLE.TYPE.CODE.1 == "Fire" ~ "Firetruck",
VEHICLE.TYPE.CODE.1 == "Firet" ~ "Firetruck",
VEHICLE.TYPE.CODE.1 == "FIRE" ~ "Firetruck",
VEHICLE.TYPE.CODE.1 == "FIRET" ~ "Firetruck",
VEHICLE.TYPE.CODE.1 == "fire" ~ "Firetruck",
VEHICLE.TYPE.CODE.1 == "FIRE" ~ "Firetruck",
VEHICLE.TYPE.CODE.1 == "EMERG" ~ "Ambulance",
VEHICLE.TYPE.CODE.1 == "AMB" ~ "Ambulance",
VEHICLE.TYPE.CODE.1 == "AMBU" ~ "Ambulance",
VEHICLE.TYPE.CODE.1 == "AMBUL" ~ "Ambulance",
VEHICLE.TYPE.CODE.1 == "ambul" ~ "Ambulance",
VEHICLE.TYPE.CODE.1 == "Amb" ~ "Ambulance",
VEHICLE.TYPE.CODE.1 == "Ambul" ~ "Ambulance",
VEHICLE.TYPE.CODE.1 == "Ambulance" ~ "Ambulance",
VEHICLE.TYPE.CODE.1 == "deliv" ~ "Delivery",
VEHICLE.TYPE.CODE.1 == "delvi" ~ "Delivery",
VEHICLE.TYPE.CODE.1 == "DELIV" ~ "Delivery",
VEHICLE.TYPE.CODE.1 == "DELV" ~ "Delivery",
VEHICLE.TYPE.CODE.1 == "Deliv" ~ "Delivery",
VEHICLE.TYPE.CODE.1 == "DUMP" ~ "Dump",
VEHICLE.TYPE.CODE.1 == "dump" ~ "Dump",
VEHICLE.TYPE.CODE.1 == "BOX T" ~ "Box Truck",
VEHICLE.TYPE.CODE.1 == "Bobca" ~ "Station Wagon/Sport Utility Vehicle",
VEHICLE.TYPE.CODE.1 == "Armored Truck" ~ "Truck",
VEHICLE.TYPE.CODE.1 == "Beverage Truck" ~ "Truck",
VEHICLE.TYPE.CODE.1 == "truck" ~ "Truck",
VEHICLE.TYPE.CODE.1 == "COM" ~ "Commercial",
VEHICLE.TYPE.CODE.1 == "Comme" ~ "Commercial",
VEHICLE.TYPE.CODE.1 == "FDNY" ~ "Firetruck",
VEHICLE.TYPE.CODE.1 == "Fdny" ~ "Firetruck",
VEHICLE.TYPE.CODE.1 == "ford" ~ "Ford",
VEHICLE.TYPE.CODE.1 == "DIRT" ~ "Station Wagon/Sport Utility Vehicle",
VEHICLE.TYPE.CODE.1 == "Unkno" ~ "Unknown",
VEHICLE.TYPE.CODE.1 == "Van Camper" ~ "Van",
VEHICLE.TYPE.CODE.1 == "moped" ~ "Moped",
VEHICLE.TYPE.CODE.1 == "GOLF" ~ "Carry All",
VEHICLE.TYPE.CODE.1 == "H1" ~ "Station Wagon/Sport Utility Vehicle",
VEHICLE.TYPE.CODE.1 == "Hears" ~ "Hearse",
VEHICLE.TYPE.CODE.1 == "Hopper" ~ "Freight",
VEHICLE.TYPE.CODE.1 == "MTA" ~ "City",
VEHICLE.TYPE.CODE.1 == "OMR" ~ "Omnibus",
VEHICLE.TYPE.CODE.1 == "OMS" ~ "Omnibus",
VEHICLE.TYPE.CODE.1 == "SCOOT" ~ "Motorscooter",
VEHICLE.TYPE.CODE.1 == "TR-Tr" ~ "Trailer",
VEHICLE.TYPE.CODE.1 == "TRAIL" ~ "Trailer",
VEHICLE.TYPE.CODE.1 == "TRL" ~ "Trailer",
VEHICLE.TYPE.CODE.1 == "Trail" ~ "Trailer",
VEHICLE.TYPE.CODE.1 == "UNK" ~ "Unknown",
VEHICLE.TYPE.CODE.1 == "US PO" ~ "USPS",
VEHICLE.TYPE.CODE.1 == "posta" ~ "USPS",
VEHICLE.TYPE.CODE.1 == "POSTA" ~ "USPS",
VEHICLE.TYPE.CODE.1 == "MAIL" ~ "USPS",
VEHICLE.TYPE.CODE.1 == "FREIG" ~ "Freight",
TRUE ~ VEHICLE.TYPE.CODE.1
))data %>%
count(VEHICLE.TYPE.CODE.1, sort = TRUE)## # A tibble: 72 x 2
## VEHICLE.TYPE.CODE.1 n
## <chr> <int>
## 1 Sedan 18037
## 2 Station Wagon/Sport Utility Vehicle 15734
## 3 Taxi 1870
## 4 Pick-up Truck 1163
## 5 Box Truck 782
## 6 Bus 675
## 7 Tractor Truck 361
## 8 Unknown 326
## 9 Van 225
## 10 Bike 222
## # … with 62 more rowsdata <- data %>%
mutate(weekday = wday(mdy(CRASH.DATE), label = TRUE)) %>%
separate(CRASH.DATE, c("month", "day", "year"), "/")
data <- data %>%
mutate(day = as.numeric(day)) %>%
mutate(month = as.numeric(month))
data <- data %>%
mutate(daytot = case_when(
month == "1" ~ day,
month == "2" ~ day + 31,
month == "3" ~ day + 60,
TRUE ~ day + 91
))
data <- data %>%
mutate(VEHICLE.TYPE.CODE.1new = fct_lump(VEHICLE.TYPE.CODE.1, n = 14, other_level = "Other"))data %>%
count(VEHICLE.TYPE.CODE.1new, sort = TRUE)## # A tibble: 15 x 2
## VEHICLE.TYPE.CODE.1new n
## <fct> <int>
## 1 Sedan 18037
## 2 Station Wagon/Sport Utility Vehicle 15734
## 3 Taxi 1870
## 4 Pick-up Truck 1163
## 5 Box Truck 782
## 6 Other 696
## 7 Bus 675
## 8 Tractor Truck 361
## 9 Unknown 326
## 10 Van 225
## 11 Bike 222
## 12 Ambulance 201
## 13 Dump 121
## 14 Motorcycle 113
## 15 Convertible 109data <- data %>%
mutate(time_of_day = case_when(
hm(CRASH.TIME) < hm("12:00") ~ "AM",
TRUE ~ "PM"
)) %>%
mutate(person_tot = NUMBER.OF.PERSONS.INJURED + NUMBER.OF.PERSONS.KILLED)Exploratory Data Analysis
data %>%
ggplot(aes(x = weekday)) +
geom_bar() +
labs(x = "Day of Week",
y = "Frequency")
data %>%
count(weekday)## # A tibble: 7 x 2
## weekday n
## <ord> <int>
## 1 Sun 4572
## 2 Mon 5826
## 3 Tue 5669
## 4 Wed 5803
## 5 Thu 6266
## 6 Fri 6835
## 7 Sat 5664data %>%
ggplot(aes(x = time_of_day)) +
geom_bar() +
labs(x = "Time of Day",
y = "Frequency")
data %>%
count(time_of_day)## # A tibble: 2 x 2
## time_of_day n
## <chr> <int>
## 1 AM 15753
## 2 PM 24882data %>%
ggplot(aes(x = person_tot)) +
geom_histogram(breaks = seq(0, 10, by = 1)) +
labs(x = "People Injured or Killed",
y = "Frequency")
data %>%
count(person_tot)## # A tibble: 11 x 2
## person_tot n
## <int> <int>
## 1 0 31645
## 2 1 7005
## 3 2 1320
## 4 3 450
## 5 4 142
## 6 5 45
## 7 6 15
## 8 7 4
## 9 8 5
## 10 9 3
## 11 10 1data %>%
summarise(mean = mean(person_tot),
sd = sd(person_tot))## mean sd
## 1 0.2948936 0.6580791data %>%
summarise(mean = mean(person_tot),
sd = sd(person_tot),
med = median(person_tot),
iqr = IQR(person_tot))## mean sd med iqr
## 1 0.2948936 0.6580791 0 0data %>%
ggplot(aes(x = month)) +
geom_bar() +
labs(x = "Month",
y = "Frequency")
### March lower, April not done yet but still proportionally even lower
data %>%
ggplot(aes(x=daytot)) +
geom_freqpoly()## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
data %>%
count(month)## # A tibble: 4 x 2
## month n
## <dbl> <int>
## 1 1 14277
## 2 2 13667
## 3 3 10967
## 4 4 1724data %>%
count(month)## # A tibble: 4 x 2
## month n
## <dbl> <int>
## 1 1 14277
## 2 2 13667
## 3 3 10967
## 4 4 1724data %>%
ggplot(aes(x = weekday, fill = time_of_day)) +
geom_bar(position = "fill") +
labs(title = "Relationship between Day of Week and Time of day",
x = "Day of week", y = "Proportion")
table <- table(data$time_of_day, data$weekday)
prop.table(table, margin = 2)##
## Sun Mon Tue Wed Thu Fri Sat
## AM 0.3589239 0.4086852 0.4065973 0.4041013 0.4052027 0.3752743 0.3490466
## PM 0.6410761 0.5913148 0.5934027 0.5958987 0.5947973 0.6247257 0.6509534data %>%
ggplot(aes(x = weekday, y = person_tot)) +
geom_boxplot() +
labs(title = "Relationship between Day of Week and Killed/Injured",
x = "day of week", y = "Killed/Injured")
data %>%
group_by(weekday) %>%
summarise(mean = mean(person_tot),
sd = sd(person_tot),
med = median(person_tot),
iqr = IQR(person_tot))## # A tibble: 7 x 5
## weekday mean sd med iqr
## <ord> <dbl> <dbl> <dbl> <dbl>
## 1 Sun 0.340 0.747 0 0
## 2 Mon 0.281 0.616 0 0
## 3 Tue 0.297 0.649 0 0
## 4 Wed 0.283 0.637 0 0
## 5 Thu 0.295 0.643 0 0
## 6 Fri 0.274 0.631 0 0
## 7 Sat 0.308 0.699 0 0data %>%
ggplot(aes(x = weekday, fill = time_of_day)) +
geom_bar(position = "fill") +
labs(title = "Relationship between Day of Week and Time of day",
x = "Day of week", y = "Proportion")
data %>%
ggplot(aes(x = month, fill = time_of_day)) +
geom_bar(position = "fill") +
labs(title = "Relationship between Month and Time of day",
x = "Day of week", y = "Proportion")
table <- table(data$time_of_day, data$month)
prop.table(table, margin = 2)##
## 1 2 3 4
## AM 0.3915388 0.4004536 0.3754901 0.3317865
## PM 0.6084612 0.5995464 0.6245099 0.6682135data %>%
ggplot(aes(x = month, fill = weekday)) +
geom_bar(position = "fill") +
labs(title = "Relationship between Month and Day of Week",
x = "Month", y = "Proportion")
table <- table(data$weekday, data$month)
prop.table(table, margin = 2)##
## 1 2 3 4
## Sun 0.09988093 0.11209483 0.13048236 0.10614849
## Mon 0.12376550 0.14311846 0.17005562 0.13805104
## Tue 0.12992926 0.13697227 0.15637823 0.13167053
## Wed 0.15899699 0.13141143 0.13367375 0.15719258
## Thu 0.17111438 0.15167923 0.13622686 0.14849188
## Fri 0.18043006 0.16550816 0.15573995 0.16763341
## Sat 0.13588289 0.15921563 0.11744324 0.15081206data %>%
ggplot(aes(x = as.character(month), y = person_tot)) +
geom_boxplot() +
labs(title = "Relationship between Month and Killed/Injured",
x = "month", y = "Killed/Injured")
data %>%
ggplot(aes(x = daytot, y = person_tot)) +
geom_point()
data %>%
group_by(month) %>%
summarise(mean = mean(person_tot),
sd = sd(person_tot),
med = median(person_tot),
iqr = IQR(person_tot))## # A tibble: 4 x 5
## month mean sd med iqr
## <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 1 0.296 0.659 0 0
## 2 2 0.297 0.658 0 0
## 3 3 0.287 0.658 0 0
## 4 4 0.325 0.651 0 0data %>%
ggplot(aes(x = month, fill = VEHICLE.TYPE.CODE.1new)) +
geom_bar(position = "fill") +
labs(title = "Relationship between month and vehicle",
x = "Month", y = "Proportion")
table <- table(data$VEHICLE.TYPE.CODE.1new, data$month)
prop.table(table, margin = 2)##
## 1 2 3
## Ambulance 0.004482734 0.005048657 0.004741497
## Bike 0.004342649 0.004609644 0.007203428
## Box Truck 0.019051621 0.019828785 0.020151363
## Bus 0.018421237 0.017706885 0.014315674
## Convertible 0.002241367 0.003146265 0.002370749
## Dump 0.002941794 0.002707251 0.003464940
## Motorcycle 0.001821111 0.002414575 0.003556123
## Pick-up Truck 0.028507390 0.028096876 0.030181453
## Sedan 0.436576312 0.440769737 0.448527400
## Station Wagon/Sport Utility Vehicle 0.398613154 0.382819931 0.382146439
## Taxi 0.044827345 0.053486500 0.042582292
## Tractor Truck 0.008615255 0.009292456 0.008571168
## Unknown 0.008124956 0.008048584 0.007659342
## Van 0.005113119 0.005926685 0.005744506
## Other 0.016319955 0.016097168 0.018783624
##
## 4
## Ambulance 0.009280742
## Bike 0.010440835
## Box Truck 0.010440835
## Bus 0.007540603
## Convertible 0.004640371
## Dump 0.002320186
## Motorcycle 0.008700696
## Pick-up Truck 0.023781903
## Sedan 0.499419954
## Station Wagon/Sport Utility Vehicle 0.359628770
## Taxi 0.018561485
## Tractor Truck 0.009860789
## Unknown 0.009280742
## Van 0.004640371
## Other 0.021461717The above plot demonstrates the preliminary exploratory data analysis for our research.
The frequency of over the weekdays seem to be evenly distributed. However, weekend crashes differ: Fridays seem to have the most number of crashes, whereas Saturdays and Sundays (especially Sundays) seem to result in far fewer crashes. There seem to be considerably more crashes that occur in PM hours than AM hours. The vast majority of crashes result in 0 injuries and fatalities. The histogram for the number of total injuries and fatalities is highly skewed right. The average number of injuries/fatalities per crash is .294 with a standard deviation of .658, whereas both the median and IQR are 0. January and February seem to have an almost equal number of crashes. March has considerably less crashes than both January or February, and April has considerably less crashes than March, even considering the fact that approximately only half of April’s total crashes are represented, as the dataset only records observations through April 14th. The frequency polygraph graphs the number of crashes per day against time. We can see a constant fluctuation around 1750 throughout March, which starts to drop signficantly around day 70 (which is around March 10th). Most of the vehicle types recorded are Sedans, followed by Sports cars.
Comparing the day of week and time of day for the crashes, the weekend days seem to have a slightly higher proportion of PM crashes than do the weekday days. There doesn’t seem to be much relationship between day of week and total number of injuries and fatalities for the crashes, although it is hard to tell because of the extreme skewness of the plot. Summary statistics reveal slightly higher injury/fatality rates for Saturdays and Sundays (although the median and IQR are 0 for each day of the week), but we don’t know yet if this is significant. There seem to be a higher proportion in PM crashes instead of AM crashes in March and April, compared to January and February. There doesn’t seem to be much difference in the distribution of the days of the week of the crashes over the four months. Comparing month of crashes and number of injuries/fatalities per crash, there seem to be less crashes in April with a higher number of injuries/fatalities, but it is important to keep in mind that only half of April’s days are observed. However, summary statistics demonstrate that there doesn’t seem to be much significant difference in the average number of people injured/killed per crash, or the standard deviation (April has a slightly higher average, but especially because it has less datapoints, we would need to do more tests to determine the significance of this difference.) Graphing the number of fatalities/injuries in each crash over time, we can wee that the number of injuries and fatalities in each crash have much less scatter after around day 74 (about March 14th), which demonstrates that there have been virtually no crashes with a considerably high number (5+) of injuries/fatalities since then. Comparing month with type of vehicle, there seems to be a slightly higher proportion of Sedans in crashes for March and more so April, though it is of note that there are fewer April dates.
The following plot shows the geographic distribution of all NYC car crashes in 2020 from the dataset.
library(naniar)
library(ggmap)## Google's Terms of Service: https://cloud.google.com/maps-platform/terms/.## Please cite ggmap if you use it! See citation("ggmap") for details.data <- data %>%
replace_with_na_at(.vars = c("LONGITUDE","LATITUDE"),
condition = ~.x == 0)
qmplot(LONGITUDE, LATITUDE, data = data, color = I("red"), alpha = I(.1), size = I(.01))## Using zoom = 11...## Source : http://tile.stamen.com/terrain/11/601/768.png## Source : http://tile.stamen.com/terrain/11/602/768.png## Source : http://tile.stamen.com/terrain/11/603/768.png## Source : http://tile.stamen.com/terrain/11/604/768.png## Source : http://tile.stamen.com/terrain/11/601/769.png## Source : http://tile.stamen.com/terrain/11/602/769.png## Source : http://tile.stamen.com/terrain/11/603/769.png## Source : http://tile.stamen.com/terrain/11/604/769.png## Source : http://tile.stamen.com/terrain/11/601/770.png## Source : http://tile.stamen.com/terrain/11/602/770.png## Source : http://tile.stamen.com/terrain/11/603/770.png## Source : http://tile.stamen.com/terrain/11/604/770.png## Source : http://tile.stamen.com/terrain/11/601/771.png## Source : http://tile.stamen.com/terrain/11/602/771.png## Source : http://tile.stamen.com/terrain/11/603/771.png## Source : http://tile.stamen.com/terrain/11/604/771.png## Warning: Removed 3263 rows containing missing values (geom_point).
To try to spot patterns in scatter over time, we divide the dataset into monthly periods, and plot the geographic distribution by month.
data.s1 <- data[data$month == 1 ,]
data.s2 <- data[data$month == 2 ,]
data.s3 <- data[data$month == 3 ,]
data.s4 <- data[data$month == 4 ,]
qmplot(LONGITUDE, LATITUDE, data = data.s1, color = I("red"), alpha = I(.2), size = I(.01))## Using zoom = 11...## Warning: Removed 1124 rows containing missing values (geom_point).
qmplot(LONGITUDE, LATITUDE, data = data.s2, color = I("red"), alpha = I(.2), size = I(.01))## Using zoom = 11...## Warning: Removed 1112 rows containing missing values (geom_point).
qmplot(LONGITUDE, LATITUDE, data = data.s3, color = I("red"), alpha = I(.2), size = I(.01))## Using zoom = 11...## Warning: Removed 882 rows containing missing values (geom_point).
qmplot(LONGITUDE, LATITUDE, data = data.s4, color = I("red"), alpha = I(.2), size = I(.01))## Using zoom = 11...## Warning: Removed 145 rows containing missing values (geom_point).
From the above plots, we can see that the March and April have far less density of crashes in all of the city as a whole. We note, however, that only half of April is accounted for in the dataset. To account for this, we decrease the transparency of points and plot April again to see if there are notable differences in geographic location.
qmplot(LONGITUDE, LATITUDE, data = data.s4, color = I("red"), alpha = I(.6), size = I(.01))## Using zoom = 11...## Warning: Removed 145 rows containing missing values (geom_point).
Whereas the geographic scatterplots for January through March showed a higher concentration of crashes in the lower Manhattan area, the plot for April so far seems to show crashes that are much more evenly distributed throughout New York City (a trend that we can see starting in March, though to a lesser extent). In fact, the higher concentrated areas now seem to be the area in between Manhattan and Bronx, as well as the middle of Brooklyn. However, we must keep in mind that the April plot is constructed with fewer data points- both because of the shorter time period, and because of the overall lower frequency of crashes. Furthermore, we must keep in mind that there are several missing values that weren’t included in these geographic scatterplots, which may be missing systematically.